Biomedical Evaluation on PubMedQA and MedQA (test)
56.4PubMedQA ScoreLED
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LEDMethod Class=Model Merging2026.01 | 56.4 | 40.06 | |
| BiomedicineModel Type=Domain Expert (Meditron3-Qwen2.5-7B)2026.01 | 51 | 54.46 | |
| ReasonAnyMethod Class=Model Merging2026.01 | 49.6 | 47.96 | |
| ReasoningModel Type=Domain Expert (DeepSeek-R1-Distill-Qwen-7B)2026.01 | 38 | 30.2 | |
| DAREMethod Class=Model Merging2026.01 | 23 | 43.12 | |
| Task ArithmeticMethod Class=Model Merging2026.01 | 22.8 | 33.3 | |
| FuseLLMMethod Class=Model Merging2026.01 | 21 | 15.8 | |
| LinearMethod Class=Model Merging2026.01 | 14.6 | 33.36 | |
| TIESMethod Class=Model Merging2026.01 | 11 | 30.76 |